A robot that can think and execute instructions like a human brain has been developed to successfully navigate a maze by walking 16 steps

 　　Is it possible to implant human "intelligence" into machines? Further breakthroughs in physical-reservoir computing, a technique for understanding brain signals, have helped invent artificially intelligent robots that think like humans.

　　Recently, researchers at the Eindhoven University of Technology and the Max Planck Institute for Polymer Research at the University of Mainz have demonstrated that this vision can be realized.

　　They designed an artificial intelligence robot that can think and execute instructions like the human brain, and can smoothly navigate a maze constructed of multiple black honeycomb-shaped hexagons.

　　On December 10, 2021, a related paper was published under the title "Organic neuromorphic electronics for sensorimotor integration and learning in robotics" by Imke Klau, Ph.D. researcher at the Department of Mechanical Engineering, Eindhoven University of Technology Howson is the first author.

　　The artificial intelligence robot studied by the team has two wheels and several reflex and touch sensors, and it receives an electrical stimulus for each correct turn in the maze, which is similar to the logic of walking in a maze. It's just that mice are tuned through synapses in the brain, a breakthrough that paves the way for new applications of neuromorphic devices in clinical diagnostics.

Robot walks 16 steps to traverse a honeycomb maze made of hexagons

　　Whether the robot can successfully walk out of the maze is related to the neurons in the cell culture medium, which are the "bridge" between information flow and synapse communication. Synapses strengthen as information flows through neurons, and this plasticity functions as learning and memory, serving as a physical repository for computers to build coherent signals.

　　Normally, the neurons in the cell culture medium act as a baseline for the robot to walk in the maze, and when the robot reaches a dead end or deviates from the walker's route and the maze exit, a visual indicator conveys a "return" or "left" to the robot. "Turn" signal, the robot would keep trial and error until it got out of the maze as the electrical impulses interfered with the neurons in the cell culture.

　　Experiments show that the robot can run out after walking 16 steps in the maze formed by hexagons. It is worth noting that once it learns to walk from a specific route, it can walk out of the maze from any other route. This feature of the robot is reproducible and extensible, and can be extended to more application fields.

　　According to Imke Clawhausen, "The excellent performance of the robot in learning and out of the maze is related to the unique integration of sensors and motors."

Smart polymers are also the key to the robot's smooth escape from the maze

　　Neuronal computation can be directly simulated in the numerical domain of neuromorphic circuits, providing real-time communication between virtual worlds. Accessed through sensorimotor systems and the digital unit of robotic platforms, but these neuromorphic circuits are typically larger in scale and can only be implemented in custom robotic systems.

　　As emerging materials and devices have the ability to directly mimic biologically inspired and related functions, such as synaptic plasticity, neuronal function, homeostasis, and self-healing capabilities, without the need for complex circuits. Therefore, it can unlock circuit functions that cannot be achieved by traditional electronics.

　　Furthermore, physical bodies are crucial in robotics, for example, using inertia for energy saving and morphological adaptation, for motion in unstructured or complex environments. Until recently, small circuits based on metal oxide neuromorphic devices have been used for local computation and control of robotic systems.

　　Currently, organic electronic materials for neuromorphic electronics exhibit excellent tunability, high stability, soft organic materials that can be solution-processed, or printed with relatively low thermal budgets, and can be integrated over large areas.

　　Another highlight of the research is the use of organic materials for neuromorphic robots. In addition to being stable, this polymer can tune most of the specific states in the maze. If you take advantage of this function, you can remember certain actions or events just like neurons and synapses in the brain.

　　Compared to inorganic materials, the polymers have a greater range of tuning and the ability to "remember" or store learned states for long periods of time. Polymeric materials can also be used in many biomedical applications.

　　Based on the properties of organic materials, actual nerve cells could theoretically be integrated with smart electronic devices, says Imke Clawhausen: "If your arm is lost in an accident, these AI devices could use a bionic hand and your body Connect.” Organic neuromorphic computing can also leverage small edge computing devices to process sensor data outside the cloud.

　　The research team was asked, "Can neuromorphic robots one day be able to participate in football games like football robots?"

　　Imke Clawhausen said: "It is theoretically possible, but it will take a lot of time. For a long time.”

　　At present, the robot studied by the team still needs to be controlled by traditional software to walk around. In order for the neuromorphic robot to perform complex tasks smoothly, the next step for the team is to build a neuromorphic network that allows multiple devices to work together in a grid.